190 likes | 394 Views
Delay Tolerant Network Routing. Sathya Narayanan, Ph.D. Computer Science and Information Technology Program California State University, Monterey Bay. This work is supported by the Naval Postgraduate School – Military Wireless Communications Research Group. Overview of Talk. Background
E N D
Delay Tolerant Network Routing Sathya Narayanan, Ph.D. Computer Science and Information Technology Program California State University, Monterey Bay This work is supported by the Naval Postgraduate School – Military Wireless Communications Research Group
Overview of Talk • Background • Research Objective • Performance analysis • Message Prioritization • Simulation Study • Results • Future Plans
Delay Tolerant Network Routing • Traditional networks • Route from source to destination exists when the message leaves the source • Delay tolerant networks • No pre-existing route • Message is forwarded as nodes encounter each other • Message traverses the route over time as the nodes move around
Delay Tolerant Network D A B Node A wants to transmit to Node E E C
Delay Tolerant Network D A B Node C will encounter Node D and transmit data Data will be successfully be delivered Node A will transmit to Node B E C Node B will encounter Node C and transmit data
Routing Protocols • This research focuses on two routing protocols • Epidemic Routing • Forward message to every node encountered • Message spreads like that of a disease in a population • ProPHET • Probabilistic Routing Protocol using History of Encounters and Transitivity • Use past encounters to predict future best route • Provides a framework allowing for different forwarding decision algorithms
Research Objective • Message Prioritization • Use the insights gained from analysis to develop message prioritization algorithms for DTN routing • Performance analysis • Develop analytical and simulation models to study three related performance parameters • Duplicate messages in the network at the time of delivery • End to end latency of message delivery • Probability of message delivery
Current Status • Developed four types of ProPHET forwarding decision algorithms • Developed a simple probabilistic extension to Epidemic (q – Epidemic) • Extensive simulation analysis of Epidemic vs ProPHET routing using ONE (Opportunistic Network Environment Simulation tool)
Results • A lot of data collected • Some insights: • q = 0.5 Epidemic has similar performance as ProPHET without all the complexity when Random Waypoint Mobility is used • Aggressive algorithms have low latency at low message generation rates • We haven’t seen any consistent performance improvement by ProPHET when there is any randomness in the mobility pattern (More simulations are being run as we speak)
Results • Insights continued: • Variables that impact the latency are: • Message generation rate • Queue length • Number of nodes • Aggressive vs non aggressive algorithms
Conclusion • Throttling Epidemic behavior using a q value seems to work well • Mathematical analysis based on the input variables is needed • Work in progress • Few levers available to affect message prioritization at routing • q value for Epidemic • Limit on the number of hops • Prioritization within queues
Two Related Recent Projects • Experimentation with Simple Message Prioritization Extensions to ProPHET • NPS Master Thesis (March 2011, LT Rapin, USN) • Secure Distributed Storage for Mobile Devices • NPS Master thesis (March 2011, LT Huchton, USN) • Upcoming MILCOM paper
Experimentation with ProPHET Message Prioritization • Simple extensions (with two traffic priority classes) can increase the performance of high priority messages significantly • Higher message delivery rate • Lower message latency • Urgent need of stable software prototypes to advance DTN research beyond theory and simulations • The current IRTF DTN2 reference implementation is of very low quality
A Secure Distributed File System for Mobile Devices • Resistant to total device compromise • Up to a customizable number (k) of device captures • No need for specialized tamper-resistant hardware • Addressing limitation of “Remote Kill” • Group secret sharing also supports data resiliency • Different collection of k devices can recover data • Prototype on Android 2.2 Smart Phones • write() and read() throughput performance: up to 15 Mbps